GLP-1 Agonists and Thyroid Cancer Risk
GLP-1 Safety
1.52 odds ratio
A 2024 meta-analysis of RCTs found a modest but statistically significant increase in overall thyroid cancer risk with GLP-1 receptor agonist use.
Silverii et al., Diabetes Obes Metab, 2024
Silverii et al., Diabetes Obes Metab, 2024
View as imageEvery GLP-1 receptor agonist approved in the United States carries an FDA boxed warning about thyroid C-cell tumors. The warning exists because rodent studies showed that these drugs cause dose-dependent C-cell hyperplasia, adenomas, and carcinomas in mice and rats. Whether this finding translates to human risk has been debated since 2010, and the answer remains genuinely unresolved. For a complete overview of GLP-1 safety concerns, see our pillar article on GLP-1 drug interactions and safety.
The question matters because GLP-1 receptor agonists are now among the most widely prescribed drug classes globally. Millions of people are taking semaglutide, liraglutide, tirzepatide, and related compounds for diabetes and obesity. If there is a meaningful thyroid cancer signal, even a small increase in relative risk applied to a population this large would translate to a substantial absolute number of cases. If there is not, the boxed warning may be causing unnecessary anxiety and potentially deterring people from effective treatments.
Key Takeaways
- GLP-1 receptor agonists cause thyroid C-cell hyperplasia and medullary thyroid carcinoma in rodents at clinically relevant exposure multiples, in a dose-dependent and duration-dependent manner (Bjerre Knudsen et al., 2010)
- The rodent effect is species-specific: rodent thyroid C-cells express substantially higher GLP-1 receptor density than human C-cells, and human calcitonin levels do not increase during GLP-1 agonist treatment (Bjerre Knudsen et al., 2010)
- A 2024 meta-analysis of RCTs found a statistically significant increase in overall thyroid cancer risk (OR 1.52, 95% CI 1.01-2.29) but acknowledged the absolute risk increase is small (Silverii et al., 2024)
- Large cardiovascular outcome trials (SUSTAIN-6, LEADER) with thousands of patients showed similar thyroid cancer incidence in GLP-1 and placebo groups (Marso et al., 2016)
- The FDA boxed warning contraindicates GLP-1 agonists in patients with personal or family history of medullary thyroid carcinoma or MEN2 syndrome
The Rodent Evidence: What the Animal Data Shows
The thyroid concern originates from a single, well-conducted study. Bjerre Knudsen et al. (2010) tested liraglutide and exenatide in mice and rats at doses ranging from clinically relevant to supratherapeutic levels for up to two years. Both compounds caused dose-dependent increases in thyroid C-cell hyperplasia, C-cell adenomas, and C-cell carcinomas (medullary thyroid carcinoma, MTC) in both species.[1]
The mechanism was direct: GLP-1 receptor agonists activated GLP-1 receptors on thyroid C-cells, stimulating calcitonin release and driving C-cell proliferation. The effect was dose-dependent and duration-dependent. At the highest doses, frank MTC developed in a significant proportion of animals. At lower, clinically relevant exposures, the effect was limited to C-cell hyperplasia without progression to malignancy in most cases.
The researchers also identified a critical species difference. Rodent thyroid C-cells express substantially higher densities of GLP-1 receptors compared to human C-cells. In humans, GLP-1 receptor expression on C-cells is present but at much lower levels, estimated at no more than 30% of thyrocytes and parafollicular cells. This difference in receptor density may explain why rodent C-cells respond proliferatively to GLP-1 agonists while human C-cells may not.[1]
The calcitonin data supports this species distinction. In rodent studies, GLP-1 agonists reliably increased serum calcitonin, a marker of C-cell activity. In human clinical trials, calcitonin levels have not increased during GLP-1 agonist treatment across multiple large studies spanning years of follow-up. This absence of a calcitonin signal in humans suggests that the C-cell stimulation seen in rodents does not occur at a clinically detectable level in people.
What the Large Clinical Trials Show
The largest human safety datasets come from cardiovascular outcome trials that enrolled thousands of patients and followed them for years. These trials were not designed specifically to detect thyroid cancer (which would require much larger populations given MTC's rarity), but they provide the best available controlled human data.
The LEADER trial (Marso et al., 2016) randomized 9,340 patients with type 2 diabetes to liraglutide or placebo and followed them for a median of 3.8 years. Thyroid cancer events were similar between groups.[2] The SUSTAIN-6 trial (Marso et al., 2016) randomized 3,297 patients to semaglutide or placebo for 104 weeks and similarly found no excess thyroid malignancy signal.[3]
Husain et al. (2020) pooled safety data across the SUSTAIN and PIONEER semaglutide trial programs, encompassing both injectable and oral formulations across thousands of patient-years of exposure. Their analysis did not identify an increased thyroid cancer risk signal.[4] Peter et al. (2020) reviewed the safety profile of injectable semaglutide across completed trials and reached the same conclusion.[5]
These trials have a significant limitation: medullary thyroid carcinoma is rare in the general population, occurring at a rate of approximately 0.2 per 100,000 person-years. Even trials with 9,000+ participants followed for 3-4 years are underpowered to detect a modest increase in such a rare event. The absence of a signal in these trials is reassuring but does not constitute proof that no risk exists.
The Meta-Analysis Conflict
When individual trials are too small to detect rare events, meta-analysis pools them to increase statistical power. This is where the thyroid cancer question gets complicated, because meta-analyses have produced conflicting results.
Silverii et al. (2024) conducted a systematic review and meta-analysis of randomized controlled trials examining GLP-1 receptor agonist use and thyroid cancer risk. They found a statistically significant increase in overall thyroid cancer risk (Mantel-Haenszel odds ratio 1.52, 95% CI 1.01-2.29, P = 0.04).[6]
This is a moderate increase in relative risk. However, the absolute risk increase is very small because thyroid cancer (particularly MTC) is rare to begin with. A 52% relative increase applied to a baseline rate of approximately 0.2 per 100,000 person-years still produces a very low absolute event rate. The confidence interval barely crosses 1.0, meaning the statistical significance is borderline.
An earlier meta-analysis using data from 45 RCTs found no statistically significant effect (RR 1.30, 95% CI 0.86-1.97). The difference between these meta-analyses reflects differences in included trials, statistical methods, and the specific cancer endpoints examined. When the effect is small and the event rate is low, modest differences in methodology can push results above or below the threshold of statistical significance.
Large observational cohort studies using population-level databases have generally found no increased risk. A multi-site international cohort study pooling data from six countries (Canada, Denmark, Norway, South Korea, Sweden, and Taiwan) found no increased thyroid cancer risk with GLP-1 agonist use compared to DPP-4 inhibitor use (pooled weighted HR 0.81, 95% CI 0.59-1.12).
Why the Species Difference Matters
The core question is whether the rodent finding is relevant to humans. Several lines of evidence suggest it may not be.
First, GLP-1 receptor density on C-cells differs markedly between species. Rodent C-cells constitute a larger proportion of thyroid mass and express more GLP-1 receptors per cell than human C-cells. This means the pharmacological exposure that drives proliferation in rodents may not achieve the same receptor occupancy threshold in humans.
Second, calcitonin responses differ. GLP-1 agonists reliably increase calcitonin in rodents but not in humans across multiple trials with years of exposure. If GLP-1 agonists were stimulating human C-cells to the degree seen in rodents, calcitonin should rise. It does not.
Third, the baseline biology differs. Rodent C-cells are inherently more proliferative than human C-cells and more susceptible to pharmacologically induced hyperplasia. This is not unique to GLP-1 agonists; other compounds that cause C-cell tumors in rodents (e.g., proton pump inhibitors, calcitonin itself) have not produced corresponding thyroid cancer signals in humans.
These differences do not guarantee safety. A compound could theoretically affect human C-cells through a mechanism that does not produce a calcitonin increase or that operates below the detection threshold of current clinical monitoring. The species difference argument is strong but not conclusive.
The GLP-1 Receptor Distribution Question
Knudsen et al. (2016) demonstrated that long-acting GLP-1 receptor agonists like liraglutide directly access the hypothalamic arcuate nucleus and activate POMC/CART neurons.[7] This research on GLP-1 receptor distribution is relevant to the thyroid question because it demonstrates that these drugs reach and activate targets beyond the gut and pancreas.
The same principle applies to the thyroid. GLP-1 receptors are expressed on human thyroid tissue, albeit at lower densities than in rodents. Whether long-acting formulations (which maintain sustained plasma levels) produce more C-cell exposure than short-acting formulations is a theoretical concern that has not been systematically studied. For context on how different GLP-1 agonists compare in their pharmacological profiles, see every GLP-1 receptor agonist compared.
What the FDA Warning Means in Practice
The FDA boxed warning states that GLP-1 receptor agonists are contraindicated in patients with a personal or family history of medullary thyroid carcinoma or multiple endocrine neoplasia syndrome type 2 (MEN2). This is a precautionary restriction based on the rodent data, not on confirmed human cases.
The warning does not state that GLP-1 agonists cause thyroid cancer in humans. It states that the risk cannot be excluded based on available data. This is a regulatory distinction that reflects the uncertainty. When a drug causes cancer in rodents and the human relevance is unknown, the FDA typically applies a warning rather than waiting for definitive human evidence, which could take decades to accumulate.
For the general population without MTC risk factors, the current evidence suggests the absolute risk, if it exists, is very small. The pancreatitis concern and gastroparesis risk with GLP-1 agonists may represent more clinically relevant safety considerations for most patients, based on the current strength of evidence.
Where the Evidence Stands
The thyroid cancer question for GLP-1 agonists sits in a genuinely ambiguous space. The rodent data is clear and reproducible. The human data is reassuring but limited by statistical power for rare events. Meta-analyses disagree. Observational studies trend negative but have their own biases.
The honest summary: there may be a small increase in thyroid cancer risk with GLP-1 agonist use, or there may not. If it exists, the absolute magnitude is likely very small relative to the metabolic benefits these drugs provide for obesity and type 2 diabetes. The contraindication in patients with MTC history or MEN2 is well-justified on a precautionary basis regardless of whether the rodent finding translates to humans.
What would resolve the question is a large, long-term registry study specifically designed to track thyroid cancer incidence in GLP-1 agonist users versus matched non-users, with sufficient follow-up time (10+ years) and population size (hundreds of thousands) to detect a rare event. Several such registries are underway. Until their data matures, the answer remains uncertain.
It is also worth noting that thyroid cancer in general is subject to significant detection bias. Thyroid ultrasound screening has increased the detection of subclinical thyroid cancers that might never have caused symptoms. If GLP-1 agonist users receive more medical monitoring than the general population (which they do, given the underlying diabetes or obesity management), they may have more thyroid cancers detected through incidental findings rather than increased incidence. Separating true incidence increases from detection bias requires careful study design that most existing analyses have not fully addressed.
The distinction between thyroid cancer subtypes also matters. Papillary thyroid carcinoma (PTC) is the most common type and has an excellent prognosis. Medullary thyroid carcinoma (MTC), the type seen in rodent studies, is much rarer and more aggressive. Most of the signal in meta-analyses comes from overall thyroid cancer (predominantly PTC), not MTC specifically. Whether GLP-1 agonists affect MTC risk specifically in humans remains unknown because the event rate is too low to detect with current study sizes.
The Bottom Line
GLP-1 receptor agonists cause thyroid C-cell tumors in rodents through direct GLP-1 receptor activation on C-cells, but the human relevance of this finding is unclear. Human C-cells express substantially fewer GLP-1 receptors, calcitonin levels do not increase during treatment, and large clinical trials have not detected excess thyroid cancer. A 2024 RCT meta-analysis found a borderline-significant 52% relative increase in overall thyroid cancer risk, but the absolute risk increase is very small. The FDA boxed warning is precautionary, reflecting uncertainty rather than confirmed human harm.